I discovered last week that Fable 5 can write perfect xTensa LX7 assembler code without tools or references. Mind blown.
But, when working on a creative graphics task, the results were best in Lua, middling in integer-only C, and underwhelming in ASM in terms of creative depth.
Cool line of questioning, but one piece of information is pivotal and critically not-yet-included: equivalent accomplishments in each language. For example, if I want to write standard things: web server, memoized fibonnaci, recipe search engine, what's the length-and-density of these outputs for each language? I think that would add in some ~normalization.
> Dynamically typed languages generally have a lower LLM token cost than traditional statically typed languages because omitting explicit type declarations makes the code more compact.
If this was true, the programming languages that are very much on the left side of
should be very ideal for LLMs, in particular if they are dynamically typed.
What I can tell you is: I experimented with AI prompts for generating Wolfram (Mathematica) code using some LLMs, and I can tell you that the results were very disappointing: in my experience LLMs have difficulties with programming languages that are
- very concise, and
- for which there is less code publicly available.
Wolfram (Mathematica) is a good example of such a programming language.
I’ve actually found Sol delivers great results with Odin, despite there not being much Odin code available. I think it is able to work well with it because:
- It is a rather simple language
- It has a lot of very useful libraries already built in.
With just a single main.odin file you can do a heck of a lot stuff, which LLMs seem to like.
> I think it is able to work well with it because:
> - It is a rather simple language - It has a lot of very useful libraries already built in.
> With just a single main.odin file you can do a heck of a lot stuff, which LLMs seem to like.
Also Wolfram/Mathematica has an insane amount of useful libraries already built in (there even exists the saying "Python is 'batteries included', Wolfram is 'spaceship included'"), and also there in a single file you can do a heck of a lot stuff.
On the other hand:
- LLMs tend to hallucinate non-existing function when you ask an LLM to code something in Wolfram that is not commonly done (concerning this point, nevertheless keep in mind that Wolfram is often used for "one-of-a-kind programs", i.e. for writing very specialized programs that have possibly never been done before).
- Wolfram code tends to be quite dense.
- If there is a small mistake in Wolfram code, the code typically simply won't work.
One thing worth noting is that syntactic density doesn't necessarily mean cheaper because because symbols don't chunk/tokenize as well as plain English
What I see from results like this is that the delta between languages is small enough now that it's hard to justify not not using something like Rust for the performance and correctness benefits if you're using LLMs and it fits the domain
I love this take because I had exactly the opposite idea. I thought the combo of remarkably simple text (not even wrapped) with incredible, full width visualizations was chef's kiss. I really like the balance there personally, but I hear you. Does your browser have Reader Mode or something like that? I don't use those tools personally, but I believe they will recast the text parts into something that renders optimally for reading (ideal font size, number of characters per line, etc....).
Is doing something for yourself really working for free? That's an interesting take. But I can understand why you don't want this for yourself, so enjoy the page in all its splendour as it is already!
I mean, if it really bothers you you could fairly trivially apply picocss or whatever to it using a user stylesheet. That is so little effort that calling it working for free would be disingenuous to say the least.
I discovered last week that Fable 5 can write perfect xTensa LX7 assembler code without tools or references. Mind blown.
But, when working on a creative graphics task, the results were best in Lua, middling in integer-only C, and underwhelming in ASM in terms of creative depth.
Related:
Which programming languages are most token-efficient? - https://news.ycombinator.com/item?id=46582728 - Jan 2026 (91 comments)
Cool line of questioning, but one piece of information is pivotal and critically not-yet-included: equivalent accomplishments in each language. For example, if I want to write standard things: web server, memoized fibonnaci, recipe search engine, what's the length-and-density of these outputs for each language? I think that would add in some ~normalization.
> Dynamically typed languages generally have a lower LLM token cost than traditional statically typed languages because omitting explicit type declarations makes the code more compact.
If this was true, the programming languages that are very much on the left side of
> https://danuker.go.ro/programming-languages.html#non-math-ma...
> https://danuker.go.ro/programming-languages.html#overall-map
should be very ideal for LLMs, in particular if they are dynamically typed.
What I can tell you is: I experimented with AI prompts for generating Wolfram (Mathematica) code using some LLMs, and I can tell you that the results were very disappointing: in my experience LLMs have difficulties with programming languages that are
- very concise, and
- for which there is less code publicly available.
Wolfram (Mathematica) is a good example of such a programming language.
I’ve actually found Sol delivers great results with Odin, despite there not being much Odin code available. I think it is able to work well with it because:
- It is a rather simple language - It has a lot of very useful libraries already built in.
With just a single main.odin file you can do a heck of a lot stuff, which LLMs seem to like.
> I think it is able to work well with it because:
> - It is a rather simple language - It has a lot of very useful libraries already built in.
> With just a single main.odin file you can do a heck of a lot stuff, which LLMs seem to like.
Also Wolfram/Mathematica has an insane amount of useful libraries already built in (there even exists the saying "Python is 'batteries included', Wolfram is 'spaceship included'"), and also there in a single file you can do a heck of a lot stuff.
On the other hand:
- LLMs tend to hallucinate non-existing function when you ask an LLM to code something in Wolfram that is not commonly done (concerning this point, nevertheless keep in mind that Wolfram is often used for "one-of-a-kind programs", i.e. for writing very specialized programs that have possibly never been done before).
- Wolfram code tends to be quite dense.
- If there is a small mistake in Wolfram code, the code typically simply won't work.
One thing worth noting is that syntactic density doesn't necessarily mean cheaper because because symbols don't chunk/tokenize as well as plain English
What I see from results like this is that the delta between languages is small enough now that it's hard to justify not not using something like Rust for the performance and correctness benefits if you're using LLMs and it fits the domain
I love Dan's writing. I really do. But I don't understand why he doesn't have some basic styling on his blog so that it's easier to read.
I love this take because I had exactly the opposite idea. I thought the combo of remarkably simple text (not even wrapped) with incredible, full width visualizations was chef's kiss. I really like the balance there personally, but I hear you. Does your browser have Reader Mode or something like that? I don't use those tools personally, but I believe they will recast the text parts into something that renders optimally for reading (ideal font size, number of characters per line, etc....).
It may be an artistic/engineering choice to demonstrate what minimizing bloat to an extreme degree looks like.
https://danluu.com/web-bloat/
Users being able to supply their own stylesheet is a core tenant of CSS. Go nuts and make it look however your heart desires!
Supply my own stylesheet? No thank you, I'm not here to do work for free.
Is doing something for yourself really working for free? That's an interesting take. But I can understand why you don't want this for yourself, so enjoy the page in all its splendour as it is already!
So every person who reads Dan's blog and finds the layout too dense, they should write and maintain a stylesheet for his site?
And every person globally should do this as well for any other website that doesn't have a good default reading experience?
If most readers of danluu don’t find that, then yes!
I mean, if it really bothers you you could fairly trivially apply picocss or whatever to it using a user stylesheet. That is so little effort that calling it working for free would be disingenuous to say the least.